Is there a way to implement Friedman's ARMAX model, where X is a predictor matrix. This is basically a linear model with with ARMA errors. I can see TimeSeriesModelFit has SARIMA but nor SARIMAX.
Aparently this questions was asked before without answer.
